Tarrasch is currently a free download!

Click here to download Tarrasch for Windows.

Tarrasch out of Beta status

Version 1.00a of Tarrasch has been released. Since June 11 2009 this version of Tarrasch has been available for download from the website of the world champion Rybka engine.

Tarrasch V2.0 planned

I am working on ideas for Tarrasch V2.0. I will definitely offer full .pgn editing facilities in V2.0. Tarrasch will then be suitable for use as a personal chess database. Email me with your thoughts on V2.0. At this stage I estimate V2.0 will be available in October 2009.

What does Tarrasch offer ?

Tarrasch is intended to be simpler than any other chess GUI to use. Tarrasch doesn't restrict your freedom. Move the pieces around the board, setup any position, start playing against the engine of your choice with either color from any position at any time. Change the clock settings at any time, go ahead, give yourself a nice time handicap against the engine! You can even set the clocks to be visible and running when you're not playing the engine, that way the GUI serves as a simple board, men and clocks for a game against a real human opponent.

The simplicity and freedom of Tarrasch doesn't mean there are no powerful features. Tarrasch can use any .pgn file as an opening book. Tarrasch has an 'open book' philosophy, so you can see the book moves available at any time and you can tell when the engine has played a book move. Tarrasch will even prompt you with the book moves available if you choose a non book move. This feature, inspired by Reptor (an earlier Triple Happy program - see 'What happended to Reptor ?' below) helps you memorize opening moves.

The .pgn book that comes with Tarrasch includes training positions. It is very easy to setup any of these training positions and work on your technique by practising against an engine from the training position. Tarrasch will remind you of the book move if you stray from a path stored in the .pgn book.

Tarrasch makes it very easy to create your own .pgn files, so coaches can use Tarrasch to create their own training exercises for pupils.

Tarrasch has full or partial blindfold modes, and a 'play N half moves behind' mode. These innovative training features designed to help you improve your ability to see ahead. The Tarrasch Toy Engine is an ideal opponent using these modes (you don't really want to handicap yourself when you're playing Rybka!).

Feel Free to Get Involved

If you are interested in Tarrasch, either as a user, a developer or a publisher, talk to me.

Tarrasch FAQ

How do I make .pgn files ?

Tarrasch automatically saves all your activity to a log file in standard .pgn format. Tarrasch always appends to the log file to avoid the possibility of deleting valuable work. Temporarily change the log file name to create a brand new .pgn file. For best results, always enter player names (simply click on the player name area). Tarrasch leaves some .pgn fields as '?'. To fill these in you can use the Notepad program that comes as a standard Windows accessory to edit the .pgn file.

How do I make new training material ?

Specify the book.pgn as the log file. Then enter your training positions and moves. Use the white player name to specify a label for the training position. After entering your training positions and moves, change the log file back to its original name to avoid further changes to the book. Restart Tarrasch and the book will be 'redigested' (once only) and the training positions should now be available in the training positions box when you setup a position.

What is the best way to study an opening with Tarrasch ?

Start a new log file (see earlier sections) and then play through your study material with Tarrasch. When you're done your log file can be used as a custom book. Select the custom book from the Options menu. Play against the engine and play will be directed down your study lines. Tarrasch will prompt you if you forget one of the book moves.

What is a kibitzer ?

This is a Yiddish term that's found its way into chess parlance. A kibitzer is an (annoying) person who kibitzs, that is offers advice that tends to be noisy, unwanted and continual. Imagine an obnoxious older Jewish guy, probably from New York City, watching you play chess and pointing out the inadequacy of your play (in his opinion). A hint of humor is definitely implied by the use of this term in chess computing. When you start the kibitzer from the commands menu or using the control-K key combination, the currently selected engine will start to display its evaluation of the position currently on the board. Usually the engine will show the best four moves it considers are available, in order, with follow up play. A score (in units of one pawn) will also be offered. Positive scores mean white is better, at least as far as the engine is concerned.

If the engine is thinking about a move itself, it doesn't display the best four moves, instead it displays increasingly deep analysis of the best move it has currently found. This situation is distinguished on screen by using the term analysis rather than kibitzing.

When I stop the kibitzer, how come the kibitzing is still shown ?

This is my attempt to fix a usability problem. Often (in other chess programs) I have seen an intriguing piece of kibitzing/analysis on screen which I'd like to look at more. Typically the engine moves on to deeper analysis and overwrites the line I'm interested in before I've even memorized it. In Tarrasch in this situation, simply stop the kibitzer by pressing control-K again. At your leisure you can restart the kibitzer, or clear the kibitz text with a separate command provided for that purpose.

How can I download the free version of Rybka and use it with Tarrasch ?

The publishers of the superb Rybka chess engine kindly provide a free demo version of their engine. This engine is an ideal companion to Tarrasch. Since Tarrasch version Beta V0.901 there's no need to download it separately as it's included in the Tarrasch download available right here. Simply download and install Tarrasch, then select Rybka from the Options - Engine menu. Thanks to Vasik Rajlich for granting permission to include the Rybka demo. Please support the publishers of Rybka by purchasing the even stronger commercial version, which also works with Tarrasch.

Alternative Download Procedure for Experts only

Expert users who have already installed a previous version of Tarrasch might prefer to download the latest .zipped executables (700K as opposed to 3Meg plus for the full package) and use it to manually replace Tarrasch.exe and TarraschToyEngine.exe in directories C:\Program Files\Tarrasch and C:\Program Files\Tarrasch\Engines respectively. Click here to download the latest zipped Tarrasch executables.

Or click here to download the latest zipped Tarrasch GUI only.

Or click here to download the latest zipped Tarrasch toy engine only.

What Happened to Reptor ?

Reptor is no longer under active development. Tarrasch is a much more capable program that includes training features inspired by Reptor. However the Reptor memorization feature is more fully developed in Reptor itself, and I think Reptor remains a useful tool. So I will continue to make it available; Click here to download Reptor for Windows.

Tarrasch is actually an ideal tool for creating .pgn content for Reptor. My personal version of Reptor now uses the same improved board graphics as Tarrasch. This Reptor version is available to anyone on request.

Release Log

10June2009 Version 1.00a

• Bug fix - Turning off kibitz can no longer turn off real engine analysis or pondering.
• Tiny bug fix - Allow user to turn off pondering during game.
• First run after installation moves book to my documents\Tarrasch directory instead of copying book to my documents\Tarrasch directory.
• If user mixes swap sides and takeback (or goto previous move) commands, they don't end up swapping clocks with the engine.
• Clocks paused during clock dialog.

Note that the Tarrasch Toy Engine remains unchanged and labeled as Version Beta 0.905

4June2009 Version Beta 0.907

(Another) candidate V1.0 release.

• Real pondering support added - as it happens just turning the option on is not enough!
• Kibitz headline improvements - distinguish between kibitz (multi pv) and analysis (single pv). Analysis happens when the engine is working out a move to play now, or pondering in anticipation. If pondering the ponder move is shown.
• Move now command added, can be used to force engine to play a move
• Takeback command now works during engine's move.
• Takeback command now always takes back one ply.

Candidate V1.0 release.

• Hash, number of cpu cores and ponder supported in engine dialog
• Small graphics mode (40 pixel squares instead of 54) set at startup for low res screens. Allows operation down to 800x600.
• If the kibitzer runs during a game, it now shows the engine's thinking while the engine is thinking about its move. The output is scrolling depth info, not MultiPV info as appears when the human is to move.
• If popup book move lists run out of room at the bottom, they no longer are truncated, instead they go up instead of down
• Subtle graphics refinements - screen space is used better and rendering process now fully automatic - no longer involves me eyeballing pixels with MS Paint
• Use isready/readyok before position command and after stop command; Hopefully operation will be more synchronous and hence more reliable
• Bug fix (tiny), user can no longer set percentage >100 in book dialog
• Bug fix (small), better handling of setup a position that is checkmate or stalemate
• Add 'Round' tag and empty line between tags and moves to make log file more standards compliant

• Bug fix, bad engine bug since 0.903 sometimes attempts illegal moves (yikes!). Thank you Guenther RWBC on the Rybka forum.

• Bug fix, add the time increment (don't just send it to the engine!). Thank you Permanent Brain on the Rybka forum.

Many small refinements, mainly after doing more mix and match (engine/GUI) testing. This is a V1.0 candidate.

• Easier to restart a finished game (setup position -> current -> ok is no longer treated as setup position -> cancel)
• FEN notation always uses '/' not '\' avoids incompatibility with say Herman 2.4
• Use a timer to avoid hammering the wxWindows idle processing system
• Tarrasch toy engine generally reports on progress in more standards compliant way;
• It continually reports on progress in go infinite and go commands and reports useful information after a stop command.
• It measures time and nodes from the start of the go or go infinite command, not from the start of a particular level
• It also uses one less thread, and writes to stdout in the root thread, this means ChessBase 9 sees its kibitz output immediately, not after a timeout.
• Up until now Tarrasch toy engine did a very stupid thing. It would try to create the file 'C:\Documents and Settings\Bill\My Documents\Chess\wx\release\toy.log' and log info to it
• That was pretty bad, but worse was that on machines other than mine it would invariably fail (the directory wouldn't exist) and it would then spew the log info to stdout instead
• This would ensure that the engine wouldn't work with other GUIs (the only surprising thing is that it did work with my GUI!)
• Needless to say, all this stupid behaviour is now fixed.
• Upgrade to the latest Inno Setup version (5.2.3) - I am hoping this will fix the uninstall doesn't remove the C:\Program Files\Tarrasch directory problem (on Vista only?)

• Bug fix, when swapping sides, swap clocks *before* starting engine.
• Bug fix, track human_is_white properly avoids mislabeling players if Engine changed in game.
• Allow user to cancel book compilation in case it's too slow.

• Added Freeware Rybka engine and made it default engine.
• More robust anti-zombie code.
• Reduce idle calls in do nothing situations - lightens CPU load.
• Restore arrow cursor in position dialog after loading book position.
• Title is always Tarrasch Chess GUI never Bill's freeware GUI.

09Apr2009 Version Beta 0.900

• Initial version